A Study on Criteria for Determining the Number of Clusters

• 발행기관: 한국자료분석학회
• 저자명: 정광모(Kwang Mo Jeong) 성주현(Ju Hyun Sung)
• 간행물 정보: 『Journal of The Korean Data Analysis Society (JKDAS)』Vol.8 No.4, 1341~1349쪽, 전체 9쪽
• 주제분류: 자연과학 > 통계학
• 파일형태: PDF
• 발행일자: 2006.08.30

영문 초록

Determining the number of clusters is a very important issue in cluster analysis. Even though there are many criteria for the number of clusters we have no fully general and satisfactory one for every clustering method. Three main criteria used in commercial statistical packages are discussed through a small Monte Carlo study in the respect of detecting the correct number of clusters in hierarchical clustering.

목차

1. Introduction

2. Clustering Algorithms

3. Three Criteria for the Number of Clusters

4. Monte Carlo Study

5. Summary

References

키워드

hierarchical clustering cubic clustering criterion pseudo F pseudo t²